X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Flanguage%2Fstats%2Ffrequencies.q;h=93e16f7f2126bed8396fa45b6a043963734a6441;hb=7658146ef345fbddfad02e76673afbb62c9aa4d6;hp=9d90e06bd96cd63c346d5d036706a6ce18ba3799;hpb=43b1296aafe7582e7dbe6c2b6a8b478d7d9b0fcf;p=pspp-builds.git diff --git a/src/language/stats/frequencies.q b/src/language/stats/frequencies.q index 9d90e06b..93e16f7f 100644 --- a/src/language/stats/frequencies.q +++ b/src/language/stats/frequencies.q @@ -1,5 +1,5 @@ /* PSPP - a program for statistical analysis. - Copyright (C) 1997-9, 2000 Free Software Foundation, Inc. + Copyright (C) 1997-9, 2000, 2007 Free Software Foundation, Inc. This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by @@ -156,8 +156,6 @@ static void add_percentile (double x) ; static struct percentile *percentiles; static int n_percentiles; -static int implicit_50th ; - /* Groups of statistics. */ #define BI BIT_INDEX #define frq_default \ @@ -198,34 +196,16 @@ static int normal; /* FIXME */ static size_t n_variables; static const struct variable **v_variables; -/* Arenas used to store semi-permanent storage. */ -static struct pool *int_pool; /* Integer mode. */ -static struct pool *gen_pool; /* General mode. */ +/* Pools. */ +static struct pool *data_pool; /* For per-SPLIT FILE group data. */ +static struct pool *syntax_pool; /* For syntax-related data. */ /* Frequency tables. */ -/* Types of frequency tables. */ -enum - { - FRQM_GENERAL, - FRQM_INTEGER - }; - /* Entire frequency table. */ struct freq_tab { - int mode; /* FRQM_GENERAL or FRQM_INTEGER. */ - - /* General mode. */ struct hsh_table *data; /* Undifferentiated data. */ - - /* Integer mode. */ - double *vector; /* Frequencies proper. */ - int min, max; /* The boundaries of the table. */ - double out_of_range; /* Sum of weights of out-of-range values. */ - double sysmis; /* Sum of weights of SYSMIS values. */ - - /* All modes. */ struct freq *valid; /* Valid freqs. */ int n_valid; /* Number of total freqs. */ @@ -303,12 +283,12 @@ cmd_frequencies (struct lexer *lexer, struct dataset *ds) { int result; - int_pool = pool_create (); + syntax_pool = pool_create (); result = internal_cmd_frequencies (lexer, ds); - pool_destroy (int_pool); - int_pool=0; - pool_destroy (gen_pool); - gen_pool=0; + pool_destroy (syntax_pool); + syntax_pool=0; + pool_destroy (data_pool); + data_pool=0; free (v_variables); v_variables=0; return result; @@ -341,14 +321,14 @@ internal_cmd_frequencies (struct lexer *lexer, struct dataset *ds) if (cmd.a_statistics[FRQ_ST_ALL]) stats |= frq_all; if (cmd.sort != FRQ_AVALUE && cmd.sort != FRQ_DVALUE) - stats &= ~frq_median; + stats &= ~BIT_INDEX (frq_median); for (i = 0; i < frq_n_stats; i++) if (cmd.a_statistics[st_name[i].st_indx]) stats |= BIT_INDEX (i); if (stats & frq_kurt) - stats |= frq_sekurt; + stats |= BIT_INDEX (frq_sekurt); if (stats & frq_skew) - stats |= frq_seskew; + stats |= BIT_INDEX (frq_seskew); /* Calculate n_stats. */ n_stats = 0; @@ -381,7 +361,14 @@ internal_cmd_frequencies (struct lexer *lexer, struct dataset *ds) add_percentile (j / (double) cmd.n_ntiles[i]); } } - + if (stats & BIT_INDEX (frq_median)) + { + /* Treat the median as the 50% percentile. + We output it in the percentiles table as "50 (Median)." */ + add_percentile (0.5); + stats &= ~BIT_INDEX (frq_median); + n_stats--; + } /* Do it! */ input = casereader_create_filter_weight (proc_open (ds), dataset_dict (ds), @@ -524,46 +511,23 @@ calc (const struct ccase *c, const struct dataset *ds) struct var_freqs *vf = get_var_freqs (v); struct freq_tab *ft = &vf->tab; - switch (ft->mode) - { - case FRQM_GENERAL: - { - /* General mode. */ - struct freq target; - struct freq **fpp; - - target.value = (union value *) val; - fpp = (struct freq **) hsh_probe (ft->data, &target); - - if (*fpp != NULL) - (*fpp)->count += weight; - else - { - struct freq *fp = pool_alloc (gen_pool, sizeof *fp); - fp->count = weight; - fp->value = pool_clone (gen_pool, - val, - MAX (MAX_SHORT_STRING, vf->width)); - *fpp = fp; - } - } - break; - case FRQM_INTEGER: - /* Integer mode. */ - if (val->f == SYSMIS) - ft->sysmis += weight; - else if (val->f > INT_MIN+1 && val->f < INT_MAX-1) - { - int i = val->f; - if (i >= ft->min && i <= ft->max) - ft->vector[i - ft->min] += weight; - } - else - ft->out_of_range += weight; - break; - default: - NOT_REACHED (); - } + struct freq target; + struct freq **fpp; + + target.value = (union value *) val; + fpp = (struct freq **) hsh_probe (ft->data, &target); + + if (*fpp != NULL) + (*fpp)->count += weight; + else + { + struct freq *fp = pool_alloc (data_pool, sizeof *fp); + fp->count = weight; + fp->value = pool_clone (data_pool, + val, + MAX (MAX_SHORT_STRING, vf->width)); + *fpp = fp; + } } } @@ -580,27 +544,15 @@ precalc (struct casereader *input, struct dataset *ds) output_split_file_values (ds, &c); case_destroy (&c); - pool_destroy (gen_pool); - gen_pool = pool_create (); + pool_destroy (data_pool); + data_pool = pool_create (); for (i = 0; i < n_variables; i++) { const struct variable *v = v_variables[i]; struct freq_tab *ft = &get_var_freqs (v)->tab; - if (ft->mode == FRQM_GENERAL) - { - ft->data = hsh_create (16, compare_freq, hash_freq, NULL, v); - } - else - { - int j; - - for (j = (ft->max - ft->min); j >= 0; j--) - ft->vector[j] = 0.0; - ft->out_of_range = 0.0; - ft->sysmis = 0.0; - } + ft->data = hsh_create (16, compare_freq, hash_freq, NULL, v); } } @@ -729,7 +681,6 @@ postprocess_freq_tab (const struct variable *v) size_t i; ft = &get_var_freqs (v)->tab; - assert (ft->mode == FRQM_GENERAL); compare = get_freq_comparator (cmd.sort, var_get_type (v)); /* Extract data from hash table. */ @@ -777,7 +728,6 @@ static void cleanup_freq_tab (const struct variable *v) { struct freq_tab *ft = &get_var_freqs (v)->tab; - assert (ft->mode == FRQM_GENERAL); free (ft->valid); hsh_destroy (ft->data); } @@ -787,9 +737,6 @@ cleanup_freq_tab (const struct variable *v) static int frq_custom_variables (struct lexer *lexer, struct dataset *ds, struct cmd_frequencies *cmd UNUSED, void *aux UNUSED) { - int mode; - int min = 0, max = 0; - size_t old_n_variables = n_variables; size_t i; @@ -802,31 +749,6 @@ frq_custom_variables (struct lexer *lexer, struct dataset *ds, struct cmd_freque PV_APPEND | PV_NO_SCRATCH)) return 0; - if (!lex_match (lexer, '(')) - mode = FRQM_GENERAL; - else - { - mode = FRQM_INTEGER; - if (!lex_force_int (lexer)) - return 0; - min = lex_integer (lexer); - lex_get (lexer); - if (!lex_force_match (lexer, ',')) - return 0; - if (!lex_force_int (lexer)) - return 0; - max = lex_integer (lexer); - lex_get (lexer); - if (!lex_force_match (lexer, ')')) - return 0; - if (max < min) - { - msg (SE, _("Upper limit of integer mode value range must be " - "greater than lower limit.")); - return 0; - } - } - for (i = old_n_variables; i < n_variables; i++) { const struct variable *v = v_variables[i]; @@ -838,25 +760,8 @@ frq_custom_variables (struct lexer *lexer, struct dataset *ds, struct cmd_freque "subcommand."), var_get_name (v)); return 0; } - if (mode == FRQM_INTEGER && !var_is_numeric (v)) - { - msg (SE, _("Integer mode specified, but %s is not a numeric " - "variable."), var_get_name (v)); - return 0; - } - vf = var_attach_aux (v, xmalloc (sizeof *vf), var_dtor_free); - vf->tab.mode = mode; vf->tab.valid = vf->tab.missing = NULL; - if (mode == FRQM_INTEGER) - { - vf->tab.min = min; - vf->tab.max = max; - vf->tab.vector = pool_nalloc (int_pool, - max - min + 1, sizeof *vf->tab.vector); - } - else - vf->tab.vector = NULL; vf->n_groups = 0; vf->groups = NULL; vf->width = var_get_width (v); @@ -903,7 +808,7 @@ frq_custom_grouped (struct lexer *lexer, struct dataset *ds, struct cmd_frequenc if (nl >= ml) { ml += 16; - dl = pool_nrealloc (int_pool, dl, ml, sizeof *dl); + dl = pool_nrealloc (syntax_pool, dl, ml, sizeof *dl); } dl[nl++] = lex_tokval (lexer); lex_get (lexer); @@ -974,13 +879,9 @@ add_percentile (double x) if (i >= n_percentiles || x != percentiles[i].p) { - percentiles = pool_nrealloc (int_pool, percentiles, + percentiles = pool_nrealloc (syntax_pool, percentiles, n_percentiles + 1, sizeof *percentiles); - - if (i < n_percentiles) - memmove (&percentiles[i + 1], &percentiles[i], - (n_percentiles - i) * sizeof (struct percentile) ); - + insert_element (percentiles, n_percentiles, sizeof *percentiles, i); percentiles[i].p = x; n_percentiles++; } @@ -1336,28 +1237,9 @@ calc_stats (const struct variable *v, double d[frq_n_stats]) double rank; int i = 0; int idx; - double *median_value; /* Calculate percentiles. */ - /* If the 50th percentile was not explicitly requested then we must - calculate it anyway --- it's the median */ - median_value = 0 ; - for (i = 0; i < n_percentiles; i++) - { - if (percentiles[i].p == 0.5) - { - median_value = &percentiles[i].value; - break; - } - } - - if ( 0 == median_value ) - { - add_percentile (0.5); - implicit_50th = 1; - } - for (i = 0; i < n_percentiles; i++) { percentiles[i].flag = 0; @@ -1437,9 +1319,6 @@ calc_stats (const struct variable *v, double d[frq_n_stats]) percentiles[i].value = percentiles[i].x1 + ( percentiles[i].x2 - percentiles[i].x1) * s ; - - if ( percentiles[i].p == 0.50) - median_value = &percentiles[i].value; } @@ -1476,7 +1355,6 @@ calc_stats (const struct variable *v, double d[frq_n_stats]) d[frq_max] = ft->valid[ft->n_valid - 1].value[0].f; d[frq_mode] = X_mode; d[frq_range] = d[frq_max] - d[frq_min]; - d[frq_median] = *median_value; d[frq_sum] = d[frq_mean] * W; d[frq_stddev] = sqrt (d[frq_variance]); d[frq_semean] = d[frq_stddev] / sqrt (W); @@ -1493,11 +1371,6 @@ dump_statistics (const struct variable *v, int show_varname) struct tab_table *t; int i, r; - int n_explicit_percentiles = n_percentiles; - - if ( implicit_50th && n_percentiles > 0 ) - --n_percentiles; - if (var_is_alpha (v)) return; ft = &get_var_freqs (v)->tab; @@ -1509,7 +1382,7 @@ dump_statistics (const struct variable *v, int show_varname) } calc_stats (v, stat_value); - t = tab_create (3, n_stats + n_explicit_percentiles + 2, 0); + t = tab_create (3, n_stats + n_percentiles + 2, 0); tab_dim (t, tab_natural_dimensions); tab_box (t, TAL_1, TAL_1, -1, -1 , 0 , 0 , 2, tab_nr(t) - 1) ; @@ -1537,14 +1410,17 @@ dump_statistics (const struct variable *v, int show_varname) tab_float(t, 2, 1, TAB_NONE, ft->total_cases - ft->valid_cases, 11, 0); - for (i = 0; i < n_explicit_percentiles; i++, r++) + for (i = 0; i < n_percentiles; i++, r++) { if ( i == 0 ) { tab_text (t, 0, r, TAB_LEFT | TAT_TITLE, _("Percentiles")); } - tab_float (t, 1, r, TAB_LEFT, percentiles[i].p * 100, 3, 0 ); + if (percentiles[i].p == 0.5) + tab_text (t, 1, r, TAB_LEFT, _("50 (Median)")); + else + tab_float (t, 1, r, TAB_LEFT, percentiles[i].p * 100, 3, 0); tab_float (t, 2, r, TAB_NONE, percentiles[i].value, 11, 3); }